“Online platforms become the perfect intermediaries through which users can present what they want in a way they want”
~ Moga & Rughinis, 2023, p. 426
The Creation Process
In creating the complex digital self-representations of myself, it required me to first select a technology that could make those kinds of representations. I decided upon Lensa AI, based on past research into the platform and my understanding around copyright and ownership of the created images. With Lensa AI, the user maintains copyright of the images they upload and the images that are created. User images are only used to create the AI representations they request and are not used to further train AI (Prisma Labs, 2023a). Lensa AI is also a popular AI image generator and is more frequently referenced in digital communications literature.
Figure 1. Guidelines for Lensa AI photo uploads. (Prisma Labs, 2024).
To create my self-representations, I completed the following steps.
I selected 10 “good” photos of myself (Lensa required the submission of 10 to 20 photos) that featured the following characteristics:
9 selfies
Different hair styles (only one photo without fringe bangs)
Indoor and outdoor backgrounds
With and without makeup
Personal and professional settings
Smiling and not-smiling
Teeth showing and not showing
5 photos wearing large, dangling earrings
5 wearing no earrings
I used the magic avatar feature to create a package of 50 images in 10 different styles:
Quiet luxury
Retro 30s
Retro 50s
Business
Warrior princess
Flowers
Fantasy Fudo
Cyberpunk
Dark Elves
Magic School
I considered the following to narrow down my analysis to two images:
Some of the representations had artefacts and defects (5 images), including artist marks, photographer watermarks, or random or inappropriately placed strands of hair.
Resemblance to my representation in images I provided, including hair colour and style, eye colour, face shape, and expression.
Whether the Google Photos’ AI recognized the images as me
My ability to see myself represented in the photos, and not that of someone with passing resemblance to me
Analysis of Representation
Figure 3. Typology of Selfies (Zhao and Zappavigna, 2018a; Ross and Zappavigna, 2019, O’Hagan and Spilioti, 2020).
For my analysis, I chose to review my digital self-representations through a semiotic lens, supported by using O’Hagan and Spilioti’s adaptation of Zhao and Zappavigna’s (2018, 2020) typology of selfies (Figure. 3).
Semiotic communication “is a process in which language and other sign systems come to have shared meanings and thereby serve as a medium for common understanding between individuals”
~ Craig & Muller, 2007, p.163
Most of the styles I chose returned a generic dark haired woman, most often with bangs, and eyes that ranged from light blue to brown and everything in between. While many of these bore a passing resemblance to myself (dark brown hair, blue eyes, female), only a few felt like they actually looked like me. While I selected a wide range of styles to see what results I would get, I ended up selecting two self-representations that I felt best represented how I actually look and the images that I had shared - one from Quiet luxury (Figure 4) and one from Business (Figure 5). These styles were some of the most realistic images developed; the more fantastical the image style, the less I felt that it actually represented me.
Features of the images that resembled me:
bangs
larger earrings
wavy or curled hair
closed mouth smile
no freckles
eye colour ranging from blue to grey
These slight, but significant changes altered how I normally appear and present myself, both in person and online. These changes are very focused on what is popularized in an ideal appearance, likely what makes these AI self-representations so popular in the curated feeds of social media. The images I submitted lack makeup, show uneven teeth, straight hair, freckles, and feature minimal jewellery.
Incorporation of Feedback
I found the peer review process to be quite interesting. I felt the feedback I received would be helpful to strengthen my original paper in a scholarly manner, however, I struggled with how to incorporate the feedback into a blog format. The feedback I received recommended making more scholarly connections on strategic self-presentation and identity curation, deeper engagement with theoretical frameworks, incorporating more diverse and recent scholarship, and expanding upon my scholarly references. Incorporating these suggestions into my already existing paper, would strengthen it in a scholarly way, however deepening connections and addressing new sources typically lengthens academic writing, therefore I found it particularly challenging to incorporate. As most of the recommendations connected to ethical and cultural considerations, I focused on adding new sources and deepening connections in that section.
Ethical and Cultural Considerations
“Social media allows users to express, explore and experiment with identity” (O’Hagan & Spilioti, 2020, p. 1)
Tiidenberg (2018) notes how “increases in women’s gendered, embodied, and sexual visibility are seen as leading to objectification rather than increased agency, because of the existing structures of inequality” (p. 63). Societal bias is inherent in the algorithm and training datasets. “Truth…is less about a factual representation or an agreed-upon reality and more commonly about a jumble of images scraped from whatever various sources were available” (Crawford, 2021, p. 96). My selfies featured allusions towards luxury and privilege and are likely connected to what many social media users would like their images to convey. Selfies often draw upon multiple overlapping traits, such as exclusivity - the glamorisation of elite practices and possession; exotic - content removed from audiences comfort zone; exceptional - unusual abilities, qualities, or skills; and everydayness - mundane content from daily life presented with candor and insight (Dezuanni, et al., 2022, p. 358).
AI augmented selfies allow for even more creative freedom and empowerment than traditional selfies or filters. Images generated by Lensa AI can act as a way for users to display their emotions and allow users to cultivate feelings of empowerment, giving the user the freedom to post images that they might otherwise not have the courage to express (Moga and Rughinis, 2023, p. 428). However, these feelings of empowerment seem to be limited to those who feel represented by the images created - often those who fit into the dominant societal narrative. This does not account for individuals of minority groups, such as individuals of colour, women, or the LGBTQ2I+ community. It also shows preferential representation for Westernized nations, and does not make space for other cultural groups.
Transliteracy
Translating my original critical analysis into a blog format was challenging. Blogs tend to be more informal and shorter than my critical analysis was. As the feedback I received for my critical analysis made recommendations for my analysis to include more sources and make deeper connections to academic scholarship, this became increasingly challenging. I tried to focus on adapting some of my content into different modalities, such as pop-out quotes, tables, and lists. I also looked at different ways to incorporate multimedia that wouldn’t be suitable to a traditional academic paper. The most challenging section to transliterate was the Ethical and Cultural Considerations, as including the sources was vital to the connections I was trying to make, and this was the area that required me to try and incorporate the most feedback.
Awareness of ones audience and the ability for social media to travel widely, likely impacts on how I feel about my AI generated self-representations. My personal perceptions don’t align with the scope of societal pressures and expectations, particularly for women. “This ‘self-branding’ is particularly important for celebrities who, subject to greater scrutiny than the general public, undertake strategic performances designed to promote idealised versions of their self” (O’Hagan & Spilioti, 2020, p. 1). The AI self-representation algorithms curated a version of myself that doesn’t necessarily represent who I am or who I present to the world based on societal ideals and expectations, and there are other individuals that are facing misrepresentation by the biased AI algorithms. They face ongoing societal pressure from social media feeds to reflect an image that is non-representative of their image or their character.
References
Burgess, J., Leary, M., & Mate, S. (2023). AI Art: An examination of the ethics of the Lensa App. (unpublished paper for COMM 597: An Introduction to AI Systems & Ethics).
Craig, R.T., & Muller, H.L. (Eds.). (2007). Theorizing Communication: Readings Across Traditions (pp. 163-167). Sage Publications Inc.
Crawford, K. (2021). Atlas of AI, Yale University Press.
Dezuanni, M., Reddan, B., Rutherford, L., & Schoonens, A. (2022) Selfies and shelfies on #bookstagram and #booktok – social media and the mediation of Australian teen reading, Learning, Media and Technology, 47:3, 355-372, DOI: 10.1080/17439884.2022.2068575
Hunter, T. (2022). Ai selfies are flooding your feed. Here is what to know about Lensa. Washington Post. https://www.washingtonpost.com/technology/2022/12/08/lensa-ai-portraits/
Liu, F., Ford, D., Parnin, C., & Dabbish, L. (2018). Selfies as Social Movements: Influences on Participation and Perceived Impact on Stereotypes. Proceedings of the ACM on Human-Computer Interaction. 1. 10.1145/3134707. Retrieved from: https://www.researchgate.net/publication/319650312_Selfies_as_Social_Movements_Influences_on_Participation_and_Perceived_Impact_on_Stereotypes
Moga, D. A., & Rughiniş, C. (2023). Idealized Self-Presentation through Al Avatars. A Case Study of Lensa Al. 2023 24th International Conference on Control Systems and Computer Science (CSCS), Bucharest, Romania, 2023, pp. 426-430, doi: 10.1109/CSCS59211.2023.00073.
O’Hagan, L. A., & Spilioti, T. (2021). The Edwardian Selfies: A transhistorical approach to celebrity culture and pictorial bookplates. DISCOURSE CONTEXT & MEDIA, 43. https://doi-org.login.ezproxy.library.ualberta.ca/10.1016/j.dcm.2021.100522
Prisma Labs. (2023, October 25 a). Privacy Policy. Privacy Policy. https://lensa.app/privacy
Prisma Labs. (2023, October 25 b). Terms of use. Terms of Use. https://tos.lensa-ai.com/terms
Prisma Labs. (2024, February 19, 2024). Lensa AI App. https://lensa.app/
Tiidenberg, K. (2018). Visibly ageing femininities: women’s visual discourses of being over-40 and over-50 on Instagram. Feminist Media Studies, 18(1), 61–76. https://doi.org/10.1080/14680777.2018.1409988
Zhao, S., & Zappavigna, M. (2018). Beyond the self: Intersubjectivity and the social semiotic interpretation of the selfie. New Media & Society, 20(5), 1735-1754. https://doi-org.login.ezproxy.library.ualberta.ca/10.1177/1461444817706074
Zhao, S., & Zappavigna, M. (2020). Selfies and Recontextualisation: Still life self-imaging in social media. In: Miles, M., Welch, E. (eds.), Photography and its Publics. Bloomsbury Academic, London. https://www.taylorfrancis.com/chapters/edit/10.4324/9781003103721-16/selfies-recontextualization-still-life-self-imaging-social-media-michele-zappavigna-sumin-zhao
Comments
Post a Comment